Microglia mediate memory dysfunction via excitatory synaptic elimination in a fracture surgery mouse model

Abstract

Cognitive impairment is a common issue among human patients undergoing surgery, yet the neural mechanism causing this impairment remains unidentified. Surgical procedures often lead to glial cell activation and neuronal hypoexcitability, both of which are known to contribute to postoperative cognitive dysfunction (POCD). However, the role of neuron-glia crosstalk in the pathology of POCD is still unclear. Through integrated transcriptomics and proteomics analyses, we found that the complement cascades and microglial phagocytotic signaling pathways are activated in a mouse model of POCD. Following surgery, there is a significant increase in the presence of complement C3, but not C1q, in conjunction with presynaptic elements. This triggers a reduction in excitatory synapses, a decline in excitatory synaptic transmission, and subsequent memory deficits in the mouse model. By genetically knockout out C3ar1 or inhibiting p-STAT3 signaling, we successfully prevented neuronal hypoexcitability and alleviated cognitive impairment in the mouse model. Therefore, targeting the C3aR and downstream p-STAT3 signaling pathways could serve as potential therapeutic approaches for mitigating POCD.

Introduction

Postoperative cognitive dysfunction (POCD) is a worldwide health problem; the major symptom of POCD is memory deficits which greatly elevates the risk of Alzheimer’s disease (AD) development in elderly surgical patients [1,2,3]. Approximately 10–25% of major noncardiac surgery patients experience abnormal synaptic activity and memory decline after surgery [4, 5]. Increasing evidence from both patients and animal models has indicated that neuroinflammation plays a critical role in the pathogenesis and progression of this issue [6]. However, the detailed mechanism underlying POCD still needs to be uncovered. Anesthesia and surgery can induce memory impairment by directly or indirectly regulating neuronal activity. Previous studies have mostly centered on the direct effect of anesthesia and surgery on neuronal excitability in surgery mouse models [7]. Accumulating evidence also suggests that surgery itself can induce glial cell activation and inflammatory factor release, thereby indirectly regulating neuronal excitability [8, 9]. Notably, glial cell activation and neuronal activity inhibition accompanied by recognition dysfunction are the core pathological features of POCD [10, 11]. However, to date, to what extent and how abnormal neuron-glia communication contributes to memory deficits of POCD remains unclear.

The resident immune cells of the brain, known as microglia, have been shown to modulate neuronal activity via negative feedback mechanisms [12, 13]. For instance, resting microglia can be directly recruited by hyperactive neuron-derived ATP via purinergic receptor signaling; they can also convert ATP to adenosine and suppress neuronal activity via the adenosine receptor A1R [12]. In addition to directly controlling synaptic activity, microglia can also modulate neuronal activity by regulating synaptic structure [14]. A growing body of evidence suggests that microglia can sense neuronal activity and actively participate in the pruning of weak and immature synapses [14,15,16]. Multiple studies have revealed the importance of phagocytosis of microglial excitatory synapses, in the developing brain and the context of some neurodegenerative disorders [15, 17, 18]. Interestingly, microglia overactivation associated with excitatory synapse loss was also observed in a POCD mouse model [19]. However, whether and how microglia-mediated remodeling of synaptic structure is involved in the development of neuronal hypoexcitability and learning and memory deficits in POCD is not clear.

As an important component of the innate immune system [20], complement C3 plays a key role in microglia-mediated synaptic pruning [21]. Both complement C3 and C1q have been shown to tag excitatory synaptic elements for microglial engulfment in neurodegenerative disorders such as AD [18, 22, 23]. The complement cascade is often initiated by C1q, which then activates C3 and downstream C3 receptors thereby inducing synaptic engulfment by microglia [24, 25]. C3aR, a G-protein-coupled receptor primarily expressed on microglia, has been shown to participate in excitatory synapse elimination in AD [26]. The complement system is also shown to be activated in POCD model mice [19]. However, the role of the complement microglia axis in synapse pruning and cognitive dysfunction after surgery remains to be investigated.

Method details

Animals

Male C57BL/6J mice, with an age range of 12–14 weeks, were gleaned from Beijing Vital River Laboratory Animal Technology Co. Ltd (Beijing, China). C3ar^−/− mice were a kind gift from Professor Hong Zhou, School of Life Sciences, Anhui Medical University. All mice were housed in an SPF environment with a 12-h light/dark cycle and were provided restricted food and water supply. They were acclimated to this environment for at least 1 week before being randomly assigned to experimental groups. All animal experiments adhered to the Guide for the Care and Use of Laboratory Animals of Nanjing Medical University.

Anesthesia and surgery

Orthopedic surgery was performed as per the previous description [27]. Succinctly, mice were anesthetized utilizing a 5% isoflurane for induction, and anesthesia was maintained with 2% isoflurane. After shaving the hair on the right hind limb and making a 1.5 cm-long incision in the skin, the muscles were dissected, and a stainless-steel pin was placed into the tibia; osteotomy was then performed, and the wound was stapled. Local anesthetics and antibiotics were administered after surgery. The procedure lasted approximately 20 min for each mouse and was performed under sterile conditions. The animals were checked daily for signs of lameness, infection, or bleeding. To better mimic the real-world clinical setting, we only performed surgical procedures on the surgery group, while the control group did not undergo any surgical interventions.

Magnetic resonance imaging (MRI)

Herein, 9.4 Tesla (T) MRI scanner (BRUKER, Germany) performed MRI. The mice were anesthetized with 2% isoflurane. Dexmedetomidine (0.02 mg/kg) was administered by intraperitoneal injection. Respiration and heartbeat were continuously monitored during scanning. The hippocampal volume was evaluated by T2-weighted imaging utilizing parameters listed below: matrix = 256 × 256, field of view = 18 mm × 18 mm, repetition time = 4000 ms, 4 averages, echo time = 33 ms, and slice thickness = 0.5 mm. Functional data were then obtained using shot gradient echo planar imaging utilizing parameters listed below: matrix = 125 × 50, field of view = 30 mm × 12 mm, echo time = 7.88 ms, repetition time = 1,500 ms, slice thickness = 0.5 mm, and repetitions = 200.

RNA sequencing and transcriptomics

The hippocampus was harvested 3 days postoperation (control, n = 4; surgery, n = 5). TRIzol reagent (Invitrogen, CA, USA) extracted total RNA as per the manufacturer’s specifications. After ensuring sufficient purity and quantifying the RNA concentration, libraries were built with the help of the VAHTS Universal V6 RNA-seq Library Prep Kit following the manufacturer’s protocol. DESeq25 executed differential expression analyses [28]. A Q value < 0.05 and fold-change > 2 or < 0.5 denoted thresholds for remarkably differentially expressed genes (DEGs). Moreover, OE Biotech Co., Ltd. (Shanghai, China) executed transcriptome sequencing as well as analysis. The transcriptome data have been uploaded to the SRA database (PRJNA1138556).

Proteomic analysis

The hippocampus was isolated 7 days postoperation (n = 3 per group). For label-free sample preparation, SDT buffer (4% sodium dodecyl sulfate [SDS] and 100 mM Tris-HCl [pH 7.6]) was introduced to the sample, and the mixture was placed into 2 mL tubes containing quartz sand. MP Fastprep-24 Automated Homogenizer (6.0 M/S, 30 s, twice) homogenized the lysates, sonicated, proceeded to 15 min of boiling, and centrifuged at 14,000 × g for 40 min. Then 0.22-µm filters were obtained to filter the supernatant and the concentration of proteins underwent quantification via the BCA Protein Assay Kit (P0012, Beyotime). After ensuring the quality of the samples, they were examined via a nanoElute instrument (Bruker, Bremen, Germany) coupled to a timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) that possesses a CaptiveSpray source. Afterward, a C18 analytical column (25 cm × 75 μm, 1.6 μm) that has a packed emitter tip (IonOpticks, Australia) separated the peptides. An integrated column oven from Sonation GmbH (Germany) sustained the column temperatures at 50 °C. The column was made to attain a state of equilibrium utilizing four column volumes prior to loading samples in 100% buffer A (99.9% Milli-Q water, 0.1% formic acid) (these steps were done at 800 bar). The samples were separated at 300 nL/min using a linear gradient as follows: 3% buffer B (0.1% formic acid in acetonitrile) for 3 min, 3–28% buffer B for 70 min, 28–38% buffer B for 7 min, 38–100% buffer B for 5 min, and hold at 100% buffer B for 5 min.

The timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) function was controlled in parallel accumulation–serial fragmentation mode, employing these settings: mass range, 100 to 1700 m/z; 1/K0 start, 0.6 V⋅s/cm², end 1.6 V⋅s/cm²; ramp time 100 ms; lock duty cycle to 100%; capillary voltage, 1500 V; dry gas, 3 L/min; dry temp, 180 °C; 10 MS/MS scans (total cycle time 1.16 s); charge range, 0–5; active exclusion for 0.4 min; scheduling target intensity, 20,000; intensity threshold, 2500; and CID collision energy, 42 eV.

The proteomic method used was label-free quantification. Proteins with a fold change > 1.2 or < 0.83 and p-value < 0.05 (Student’s t-test) were considered differentially expressed. The proteomic analysis was implemented by GeneChem Co., Ltd. (Shanghai, China). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [29] partner repository with the dataset identifier PXD054165.

Drugs and virus

For C3aR antagonist (C3aRA) treatment, mice were injected intraperitoneally with 1 mg/kg C3aRA (SB290157 trifluoroacetate; MedChemExpress, HY-101502 A) or 0.5% dimethyl sulfoxide (vehicle) at 1 h presurgery and 2 and 4 days postsurgery. For Stattic treatment, 10 mg/kg Stattic (MedChemExpress, HY-13818) or corn oil (vehicle) was administered intraperitoneally for 1 h of presurgery and 2 and 4 days post-surgery. A lentivirus carrying a short hairpin RNA (shRNA) targeting C3ar1 was obtained from GeneChem Co., Ltd. (Shanghai, China).

Cell culture and treatment

BV2 cells were sustained in Dulbecco’s modified Eagle medium (DMEM) enriched with 10% fetal bovine serum (FBS). Utilizing previously described criteria, primary microglia were prepared [22]. Briefly, cortices together with the hippocampi were removed from newborn mouse pups in a dissection medium (Hank’s balanced salt solution rich in 10 mM HEPES, 0.6% glucose, and 1% v/v pen/strep), sliced into small pieces, and utilized 2.5% trypsin to digest them at 37 °C for 15 min. The process proceeded with adding 1 mg/mL trypsin inhibitor, centrifugation for 5 min at 1500 rpm, trituration, and resuspension in DMEM containing 10% FBS. Cells were plated in poly-D-lysine-coated T-75 flasks at a density of 50,000 cells/cm² to produce mixed glial cultures. When they reached confluency, microglia were dissociated by taping the flasks on the table, collected in medium, seeded at a density of 50,000 cells/cm², and cultured for an additional day in poly-D-lysine-coated 12-well plates for either protein/mRNA assays or coverslips staining.

In phagocytosis assays utilizing beads, aqueous green together with red fluorescent latex beads were subjected to preopsonization in FBS for 1 h at 37 °C preliminary to being diluted in DMEM. The ultimate concentrations of the beads and FBS in DMEM were 0.01% (v/v) and 0.05% (v/v), correspondingly. After treating microglia with 10 µg/mL C3 (MedChemExpress, HY-P7863) for 1 h, the C3-containing medium was substituted with DMEM that has beads, and the cells were incubated at 37 °C for 1 h. Dimethyl sulfoxide (DMSO) or 10 µM C3aRA or 2.5 µM Stattic was incorporated with the primary microglia one hour prior to the vehicle or C3 treatment. The cells were washed rigorously by the use of ice-cold phosphate-buffered saline (PBS) six times, fixed in 4% paraformaldehyde, and prepared for staining and imaging.

BV2 cells were transfected with C3ar1-targeting shRNA (shC3ar1) or control shRNA (shCon). At 72 h after lentiviral infection, the cells were taken to conduct quantitative PCR (qPCR) or immunoblotting to measure C3ar1 or C3aR expression. Furthermore, BV2 cells with C3ar1 knockdown were treated with C3 and DMEM-containing beads, as described above.

RNA extraction and qPCR

The TRIzol reagent (Vazyme Biotech, Nanjing, China) isolated total RNA, which was used for cDNA synthesis via HiScript III RT SuperMix for qPCR (Vazyme Biotech, Nanjing, China). RNA expression levels were measured using SYBR Green (Vazyme Biotech, Nanjing, China) on an Applied Biosystems machine (Waltham, MA, USA), and the 2^−ΔΔCt method computed the relative RNA expression, which underwent normalization utilizing glyceraldehyde-3-phosphate dehydrogenase. Shanghai Generay Biotech Co., Ltd. supplied all primers (Supplementary Table 1).

Immunoblotting

Hippocampal tissue or cells were extracted in radioimmunoprecipitation assay (RIPA) buffer augmented with protease as well as phosphatase inhibitors. Then, centrifuged at 12,000 rpm for 15 min with the intent of collecting the supernatant. The protein concentration was ascertained by the BCA Assay. Dilution of protein samples was achieved by 5 × SDS-polyacrylamide gel electrophoresis buffer. 30 µg of total protein was loaded in each lane, separated on a polyacrylamide gel, and transferred onto a PVDF membrane. A 5% milk in Tris-buffered saline/0.1% Tween-20 (TBST) for 1–2 h at RT blocked the membranes, then probed overnight at 4 °C with primary antibodies, washed thrice for 10 min in TBST, incubated with secondary antibody for 1–2 h at room temperature, washed thrice for 10 min in TBST, incubated with enhanced chemiluminescence solution, and developed.

Immunostaining and image analysis

Isoflurane was involved in inducing deep anesthesia in the animals. The animals were afterward transcardially perfused with 0.1 M PBS (pH 7.4) and later with 4% paraformaldehyde (Sigma‒Aldrich, US). The brains were excised and postfixed at 4 °C for 12–16 h. Subsequent to being cryoprotected in 30% sucrose, they were imbedded in an optimal cutting temperature compound. Leica cryostat (CM 3050 S) functioned in extracting coronal sections that were 30 μm in thickness and kept at -70 °C for further use.

Fixed cultured microglia or sections on coverslips were rinsed with PBS, blocked for 2 h with PBS containing 0.3% Triton X-100 and 5% normal donkey serum, incubated overnight at 4 °C with diluted primary antibodies in blocking solution, washed with PBS containing 0.3% Triton X-100, and incubated for 1–2 h at RT with secondary antibodies. Fixed cultured cells or brain sections were imaged utilizing a Leica confocal microscope (TCS SP8) and evaluated by ImageJ. Images of three equidistant sections for each animal from 3 to 4 animals per group were taken for the animal experiment and 6–9 random fields from three batches per group were taken for the cell experiment.

For analysis of synaptic markers, confocal imaging was performed with a 63× oil objective. The optical plane that exhibited the highest synaptic staining was identified from a z-stack of 5 to 7 serial optical images (1-µm step intervals) obtained over a depth of 5–7 μm. ImageJ was used to subtract the background from the fluorescence images before quantification. For each experiment, the threshold parameters used for analysis in ImageJ were uniform for all mouse groups. The average puncta density was calculated using images of five consecutive optical planes for each brain section.

Microglial engulfment and CD68 occupancy

Brain sections were coimmunostained with anti-IBA1, anti-CD68, and anti-VGluT1 antibodies. Z-stacks composed of entire microglia were acquired on a Leica confocal microscope (TCS SP8) utilizing a 63× objective. The laser power together with gain was kept consistent across all experiments. Imaris (Bitplane) analyzed all the images by developing a 3D surface rendering of each microglia, which was thresholded to guarantee that the microglial processes were reconstructed in an accurate manner and maintained consistently from this point onwards. The CD68 channel inside microglia was masked via microglial rendering, resulting in the development of a 3D representation. The CD68 volume was then utilized to mask the VGluT1 channel. VGluT1 engulfment for each microglia was computed as the total volume of VGluT1 inside the masked CD68 volume.

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining

Neuronal cell death was detected in vivo by staining 30-µm-thick, 4% paraformaldehyde-fixed, frozen hippocampal sections utilizing the TUNEL BrightRed Apoptosis Detection Kit (Vazyme, #A113-02) according to the manufacturer’s protocol.

Golgi staining

The FD Rapid GolgiStain™ Kit (FD Neurotechnologies, Inc., USA) assisted in the morphological study of the dendritic spines and neuronal dendrites. Fresh brains went through incubation utilizing equivalent volumes of solutions A and B for 2 weeks at room temperature (RT) in a bottle covered with aluminum foil and put in the dark. After 24 h of incubation, the impregnated solution was changed. After this process, the brain tissues were introduced into solution C (which was changed at the end of 24 h), incubated at RT for 3 days in the dark, sliced into 150-µm-thick sections utilizing a Leica CM3050S cryostat (Leica Microsystems, Wetzlar, Germany) at -22 °C, and stained following the manufacturer’s specifications. At 20× and 100× magnification, images were captured under an oil immersion objective (Olympus, Tokyo, Japan). The number of dendritic lengths, branches, and spine density were estimated by the Neuro J and Sholl Analysis plugins of ImageJ (NIH, Bethesda, USA).

Electrophysiology

Slice preparation

After sacrificing the mice, their whole brains were instantly removed and soaked in ice-cold artificial cerebrospinal fluid (ACSF) that constituted the following: 185 mM sucrose, 20 mM D-glucose, 26 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 1 mM CaCl₂, and 6 mM MgCl₂ (saturated with 95% O₂/5% CO₂, pH 7.4). Brain slices (350 μm thick) were cut by a vibrating microtome (VT1200, Leica Microsystems), placed in ACSF containing 20 mM D-glucose, 124 mM NaCl, 26 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 1 mM CaCl₂, and 6 mM MgCl₂ (saturated with 95% O₂/5% CO₂, pH 7.4) for 30 min at 34 °C, and given time to recover for 1 h at RT.

Spontaneous excitatory postsynaptic current (sEPSC) recording and analysis

The slices were submerged in a recording chamber that was on regular perfusion of 2 mL/min at RT with ACSF containing 124 mM NaCl, 20 mM D-glucose, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 4 mM CaCl₂, 4 mM MgCl₂, and 50 µM picrotoxin (Sigma, R284556) to suppress GABAergic synaptic activity. Whole-cell recordings in cornu ammonis 1 (CA1) pyramidal cells were performed in voltage-clamp mode (-70 mV) utilizing a Heka EPC10 amplifier (Heka Instruments). Electrodes (3–5 MΩ) for recording were filled with an internal solution comprising 125 mM CsCl, 8 mM NaCl, 10 mM HEPES, 2 mM MgATP, 0.3 mM NaGTP, 0.2 mM EGTA, and 0.1% biocytin (Sigma, B4261) (pH = 7.4). sEPSCs were initially recorded for 8 min. They were then evaluated by the Mini Analysis software (Synaptosoft). The event detection level for synaptic currents was set at 4 pA. Later, manual checking of all the events was executed.

Long-term potential (LTP) measurement

The slices were afterward changed to a recording chamber that had a glass bottom filled with ACSF comprising 20 mM glucose, 124 mM NaCl, 26 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 4 mM CaCl₂, and 4 mM MgCl₂ (saturated with 95% O₂/5% CO₂, pH 7.4) at 2 mL/min at room temperature.

The dendritic layer of CA1 neurons was involved in recording the field excitatory postsynaptic potentials (fEPSPs). This was made possible by putting an electrode in the Schaffer collateral pathway to serve as the stimulating electrode. A stimulation intensity that elicited a fEPSP that ranged between 30 and 40% of the maximum response on the stimulus-response curve was chosen. LTP was triggered by theta burst stimulation that involves six episodes that have an interval of 10 s. Each episode entailed 5 bursts at 5 Hz and each pulse constituted 5 pulses at 100 Hz. After the tetanus, the field potential was taken for 50 min. The extent of LTP was computed as the percentage change in the average fEPSP between 40 and 50 min following the commencement of LTP induction.

Behavioral tests

Morris water maze (MWM) test

This test was applied to measure spatial learning, as well as memory, as previously described [30]. Starting 1 day after anesthesia and surgery, all mice were trained for 5 days, and on each day, they were trained four times on how to swim to a concealed platform. Each mouse was given 90 s to locate the concealed platform and was given time to stay on it for 15 s before being removed. If a mouse was unable to locate the platform within the stipulated time, the mouse received guidance to the platform and was provided with a certain time to remain on it for 15 s. The quadrant containing the platform remained constant during the MWM training, but the mice were positioned in the tank from random starting points. We calculated the time each mouse took to arrive at the platform (which was denoted as escape latency) and utilized the measurement to assess learning in the mouse. We removed the platform on day six and assessed each mouse’s memory by measuring the time they took in the required quadrant, as well as the total number of platform area crossings. Afterward, we executed comparisons of the measurements of learning and memory, and swimming speeds between the mice in the control and surgery groups.

Y-maze test

This test was executed as per the previous study and was central in ascertaining the tendency of rodents to explore unfamiliar environments [31]. Rodents naturally opt to explore a new arm of the maze instead of going to a revisited arm. In this study, the Y-maze constituted three arms, namely; A, B, and C, at 120° angles to one another. Each mouse was centrally positioned at the maze and was given an allowance of 8 min to explore the maze’s three arms. The frequency of arm entries as well as triads was observed to assist in calculating the alternation percentage. If a mouse sequentially entered three distinct arms—ABC, BCA, BAC, ACB, CBA, or CAB—it was assumed to be a spontaneous alternation. The formula for computing the alternation percentage was as follows: (number of alternations / [total number of arm entries − 2]) × 100.

Location memory test (LMT)

The LMT was performed as per the previous study [32]. Briefly, 4 metallic enclosures were aligned at the four corners of a square box (45 cm × 45 cm × 45 cm). A visual cue was instilled at the south wall of the arena, and mice were put inside it through the central part of the north wall. They were exposed to a 10-minute experience trial as well as a recall trial at the end of 24 h. In the process of the experience trial, a female mouse (which was unknown to the male mice) was introduced to one enclosure out of the four metallic enclosures. The following day, it was found that the female mouse was not there(absent). The total time taken examining the four metallic enclosures was documented in both the experience and the recall trials. The percentage of time spent interacting with the female mouse was used to measure female interest (day 1) and female location memory (day 2).

Open field test

The open field test was performed as previously described [31]. Each mouse was placed in a square chamber (40 cm × 40 cm × 40 cm), and its movement over 10 min was recorded using a digital camera and analyzed using EthoVision 8.5 (Noldus). The total distance traveled and time spent in the center of the open field arena were recorded. The chamber was cleaned with 70% ethanol between each trial.

Quantification and statistical analysis

Summary data are presented as the mean ± standard error of the mean. Two-tailed Student’s t test or the Mann–Whitney test was used to test for intergroup differences. Multigroup comparisons were performed with one-way or two-way analysis of variance followed by Tukey’s post hoc test. GraphPad Prism version 9.0 was used to analyze the data and generate graphs. A value of p < 0.05 was considered to indicate statistical significance, whereas values higher than 0.05 were considered not significant.

Results

Microglia activation is associated with excitatory synapse loss in the POCD mouse model

To study whether surgery induces microglia activation and cognitive deficits, we first established a POCD mouse model using orthopedic surgery and evaluated the memory function of the model mice by employing the MWM and Y-maze tests (Fig. 1A). Using the MWM test, we affirmed that the latency to discover the platform during the training stage was longer for mice that underwent surgery. As expected, the POCD model mice took less time in the intended quadrant and crossed fewer times on the platform area following the test phase; however, the swimming speed remained unchanged between the groups (Fig. 1B and E; F (1, 90) = 23.58, P<0.0001;t = 2.925, P = 0.009; t = 2.153, P = 0.0452; t = 0.5031, P = 0.621, respectively). Moreover, the Y-maze test results also showed that the mice that underwent surgery exhibited a lower alternation percentage than the control mice (Fig. 1F; t = 2.509, P = 0.0219). These results indicate that surgery could lead to impairment in memory and cognitive malfunction in a mouse model of POCD.

Fig. 1

Microglial activation associates with synapse abnormality and memory deficits after peripheral surgery in a mouse model of POCD. (A) Diagram showing the timeline of the POCD mouse model establishment, behavior training and test, and tissue collection. (B) Comparison of the latency to find the platform during the training stage between the control and surgery groups in the Morris water maze (MWM) test (n = 10 mice per group). (C-E) Comparison of the duration (C), crossover number (D), and swimming speed (E) in the target quadrant during the MWM testing stage between the control and surgery group (n = 10 mice per group). (F) Comparison of working memory performance between the control and surgery group in the Y-maze test (n = 10 mice per group). (G) Double-immunostaining of IBA1 (green) and CD68 (red) in the hippocampus of control and model mice at different timepoints post-surgery (3 DPO and 7 DPO). Scale bars: 40 μm–10 μm. (H) Quantification of CD68 volume within microglia at different timepoints (3 DPO and 7 DPO) (n = 15–17 cells from 3 mice per group). (I) Heatmap showing the DEGs of microglia between the control and surgery group. Each column represents an individual mouse. (J) Validation of the expression levels of genes related to microglia phagocytosis using qPCR (n = 4–5 mice per group). (K-L) Immunoblot analysis of VGluT1 and VGluT2 expression levels in the control and surgery group. (K) Representative blots using 2 mice from each group. (L) Statistics of 3–6 mice per group. (M) fMRI images showing the ALFF signal at different brain regions of model mice (signal from control mice were subtracted from mice undergoing surgery); the cool (teal-blue) colors represent the decline in ALFF signal. Data in bar graph are presented as mean ± SEM. All panels, * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by two-tailed t-test, one-way or two-way ANOVA followed by Tukey’s post hoc test

We then investigated whether microglial cells are activated and participate in synaptic pruning and memory deficits utilizing the mouse model. Interestingly, we found a significant increase in microglial immunofluorescence intensity in the mouse hippocampus (CA1 region, CA3 region, and dentate gyrus) at 3 and 7 days postoperation (DPO), and the immunofluorescence intensity returned to the baseline level at 14 DPO (Figure S1A and S1B; F (4, 10) = 5.194, P = 0.0158; F (4, 10) = 11.22, P = 0.001; F (4, 10) = 5.486, P = 0.0133; respectively). Moreover, the volume of CD68⁺ microglial lysosomes was increased significantly at 3 and 7 DPO (Fig. 1G, H and F (2, 45) = 19.73, P<0.0001). These results imply that microglial cells were activated and exhibited elevated phagocytic ability after surgery. We next profiled the gene expression signature after surgery using RNA-seq and found that the genes associated with microglial activation and phagocytic function were upregulated in the hippocampus of the model mice (Fig. 1I). The expression levels of upregulated genes that participate in synaptic pruning (Cx3cr1, Fcer1g, Fcgr1, Fcgr2b, and Trem2) were further validated using qPCR (Fig. 1J; Cx3cr1: t = 2, P = 0.0287; Fcer1g: t = 5.033, P = 0.0015; Fcgr1: t = 4.9, P = 0.0018; Fcgr2b: t = 5.184, P = 0.0013; and Trem2: t = 5.392, P = 0.001). According to our transcriptome data and the latest nomenclature guidelines, we found that microglia in our surgical model align more with the characteristics of activated reactive microglia (ARM) (Figure S1C). We next examined the expression changes of excitatory synaptic components in the hippocampus of the model mice. Using immunoblotting, we found that the protein level of VGluT1 but not VGluT2 was significantly reduced in the mice that underwent surgery (Fig. 1K, L and t = 9.156, P = 0.0008; t = 0.7085, P = 0.4966; respectively). In contrast, the levels of synaptic structure proteins, including synaptophysin, SYN1, PSD-95, and Homer1, were not remarkedly different between the control and surgery groups (Figure S1D-S1G; t = 0.08893, P = 0.9309; t = 0.7831, P = 0.4517; t = 0.616, P = 0.5517; t = 1.029, P = 0.3278; respectively).

Next, we assessed neuronal activity levels in the hippocampus of THE model mice using functional magnetic resonance imaging (fMRI). We found a significant reduction in the amplitude of low-frequency fluctuations (ALFFs, Fig. 1M) and regional homogeneity signals (ReHos, Figure S1H), while the volume of the mouse hippocampus was unaffected in the model mice, indicating that neuronal excitation was decreased after peripheral surgery (representative signals from the surgery group were obtained by subtracting the signal intensity in the control group) (Figure S1I; t = 0.1349, P = 0.8957). Together, these findings infer that peripheral surgery can induce microglial activation, excitatory synapse loss, and decreased neuronal activity in the model mice.

The expression of the complement molecule C3 is elevated in the mouse hippocampus after surgery

To investigate whether complement molecules are involved in synapse elimination in the mouse model, we first evaluated the DEGs between the control and surgery groups by means of RNA-seq. At least 782 genes were altered (746 upregulated, 36 downregulated) in the hippocampus in mice that underwent surgery compared with control mice (Figure S2A). Interestingly, both Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses affirmed that complement signaling pathways like complement receptor activity and complement and coagulation cascades were significantly activated in the mouse model (Fig. 2A and B). Moreover, the WikiPathways analysis and Reactome results both indicated that microglial phagocytosis and complement activation signaling pathway activity were increased in the mouse model (Fig. 2C and S2B). Notably, the heatmap results showed that classical complement pathway-related genes such as C1q, C4b, C3, Serping1, and C3ar1 were upregulated in the mouse model (Fig. 2D), and the gene expression levels were further validated using qPCR (Fig. 2E; C1q: t = 6.514, P = 0.0003; C4b: t = 4.021, P = 0.0051; C3: t = 2.466, P = 0.0431, Serping1: t = 2.772, P = 0.0276 and C3ar1: t = 2.863, P = 0.0242).

Fig. 2

Transcriptomics and proteomics showing the complement cascade is activated in POCD mouse model. (A) GO analysis of DEGs between the control and surgery group. (B) KEGG pathway analysis of DEGs between the control and surgery group. (C) Wikipathways analysis of DEGs between the control and surgery group. (D) Heatmap showing the DEGs related to the complement cascade between the control and surgery group. Each column represents an individual mouse. (E) Validation of the DEGs in D using qPCR (n = 4–5 mice per group). (F) KEGG pathway analysis of differentially expressed proteins (DEPs) between the control and surgery group. (G-H) Immunoblot analysis of C3 and C1q expression levels at different timepoints after surgery (1, 3 and 7 DPO). (G) Representative blots using 2 mice from each group. (H) Statistics of 4 mice per group. (I) Double-immunostaining of GFAP (green) and C3 (red) in the hippocampal CA1 region from the control and surgery group (n = 3 mice per group). Scale bars: 100 μm. Data in bar graph are presented as mean ± SEM. * P < 0.05, ** P < 0. 01, *** P < 0.001, ns, not significant; by two-tailed t-test or one-way ANOVA followed by Tukey’s post hoc test

We further performed proteomics to assess protein expression levels in the mouse model at 7 DPO. In total, 5229 proteins were discovered, and 131 proteins were differentially expressed (45 upregulated, 86 downregulated in the surgery group; Figure S2C and S2D). Similar to KEGG pathway analysis of the genes identified by transcriptomic analysis, KEGG pathway analysis of the proteomics data affirmed that complement and coagulation cascade signaling pathway activity was obviously increased in the mouse model (Fig. 2F). We then investigated which complement molecules mediate excitatory synaptic pruning in model mice, immunoblotting results showed that the expression level of complement C3 but not C1q was significantly elevated at 7 DPO (Fig. 2G, H and F (3, 12) = 5.765, P = 0.0112; F (3, 12) = 0.728, P = 0.5547). Furthermore, C3 and GFAP double immunostaining revealed that the complement C3 immunofluorescence intensity was selectively elevated in astrocytes in the hippocampal CA1 region in comparison to the CA3 region or dentate gyrus (DG) at 7 DPO (Fig. 2I, S2E-S2G; t = 3.881, P = 0.0178; t = 1.7, P = 0.1644; t = 1.176, P = 0.3048, respectively). However, microglia and neurons showed almost no co-localization with C3 (Figure S2H-S2I). Together, these results indicate that complement cascade activation and microglial phagocytosis may promote synapse elimination in the mouse model.

Complement C3 promotes excitatory synapse engulfment by microglia in the mouse hippocampus after surgery

We next investigated excitatory synapse loss in POCD model mice using immunostaining. Firstly, VGluT1 and Homer1 double staining showed that excitatory synapses were markedly lost but neurons did not undergo apoptosis in the hippocampus (Fig. 3A and D, and S3; t = 2.605, P = 0.0404; t = 1.732, P = 0.134; respectively). Secondly, the alteration of excitatory synapses was largely attributed to the loss of presynaptic terminals (Fig. 3B; t = 2.639, P = 0.0386), while postsynaptic elements remained unchanged (Fig. 3C; t = 0.03567, P = 0.9727). We further determined which complement molecules mediate excitatory synapse engulfment by microglia in the model mice. We found that the immunofluorescence intensity of both C1q and C3 was significantly increased in the mouse model; however, elevated VGluT1 colocalization was only observed with C3 but not C1q in the hippocampal CA1 region at 3 DPO (Fig. 3E and H; t = 2.041, P = 0.0874; t = 4.297, P = 0.0051; respectively). Moreover, 3D reconstructions of microglia revealed a substantial surge in the volume of VGluT1-positive puncta within microglial lysosomes, indicating that microglia may actively participate in excitatory synaptic pruning in the hippocampus of the model mice (Fig. 3I, J and t = 2.375, P = 0.0218). While our western blotting result showed no significant change in synaptophysin protein level in total hippocampal tissue. However, our immunofluorescence analysis in different hippocampal subregions revealed a solely reduction in synaptophysin expression level in the CA1 (Figure S4A and S4B; t = 4.756, P = 0.0089; t = 0.01959, P = 0.9853; t = 12.76, P = 0.0002, respectively) region but not in CA3 (Figure S4C and S4D; t = 0.2707, P = 0.8; t = 0.1579, P = 0.8822; t = 0.3492, P = 0.7446, respectively) and DG (Figure S4E and S4F; t = 0.4024, P = 0.7080; t = 0.3602, P = 0.7369; t = 0.2483, P = 0.8161, respectively). These results suggest a decrease in the total presynaptic proteins in the CA1 region of the hippocampus in the surgery group mice. Combined with our transcriptome data, gene set enrichment analysis (GSEA) revealed significant enrichment of microglial activation and complement systems in the surgery group (Figure S4G and S4H). Similarly, phagocytosis and complement-mediated synaptic pruning were enriched in the surgery group (Figure S4I-S4K). Interestingly, synaptic vesicle membrane-related genes were downregulated in the surgery group (Figure S4L). Furthermore, to assess whether C3 mediates microglial phagocytosis of excitatory synapses, we evaluated the effect of purified recombinant C3 on the uptake of fluorescent latex beads by primary microglia. At 1 h after treatment with complement C3 peptides, microglia exhibited elevated phagocytic capability, as represented by more fluorescent latex beads colocalized with microglia (Fig. 3K, L and t = 4.582, P = 0.0008; t = 6.774, P < 0.0001). Taken together, these results indicate that microglia engulfment of C3-tagged excitatory synapses led to loss of excitatory synapse in the mouse model.

Fig. 3

Complement promotes microglial engulfment of excitatory synapse in mouse hippocampus after surgery. (A) Immunofluorescence detection of VGluT1 (green) & Homer1 (red) colocalized puncta in the hippocampal stratum radiatum from the control and surgery group. Scale bars: 10 μm. (B) Quantification of VGluT1 positive puncta (n = 4 mice per group). (C) Quantification of Homer1 positive puncta (n = 4 mice per group). (D) Quantification of VGluT1 & Homer1 colocalized puncta (n = 4 mice per group). (E-F) Immunofluorescence detection of VGluT1 (green) & C1q (red) colocalized puncta in the hippocampal stratum radiatum from the control and surgery group. (E) Representative images, scale bar: 20 μm and 5 μm. (F) Quantification of C1q and VGluT1 colocalized puncta (n = 4 mice per group). (G-H) Immunofluorescence detection of VGluT1 (green) & C3 (red) colocalized puncta in the hippocampal stratum radiatum from the control and surgery group. (G) Representative images, scale bar: 10 μm and 4 μm. (H) Quantification of C3 and VGluT1 colocalized puncta (n = 4 mice per group). (I-J) Triple immunofluorescence staining of IBA1 (green), CD68 (blue) and VGluT1 (red) in the hippocampal stratum radiatum from the control and surgery group. (I) Representative images and three-dimensional (3D) reconstruction. Scale bars: 10 μm and 3 μm. (J) Quantification of VGluT1 positive puncta within microglial lysosomes (n = 21–27 cells from 4 mice per group). (K-L) Effect of C3 treatment on the phagocytosis activity of microglia cells in vitro. (K) Representative immunofluorescence images showing the fluorescent beads (green) inside the primary microglial cells after C3 or PBS treatment. Scale bars: 100 μm. (L) Quantification of the fluorescent beads taken up by primary microglial cells from (K) (left: 6–7 fields from 3 batches per group; right: 35–36 cells from 3 batches per group). Data in bar graph are presented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by two-tailed t-test

Blockage of C3aR prevents microglial activation and loss of excitatory synapse in the mouse model

We next investigated whether the C3-C3aR axis also promotes excitatory synaptic pruning in the mouse model. Transcriptomics analysis suggested that the expression of complement receptors was substantially heightened in mice that underwent surgery (Fig. 4A). Immunostaining experiments also showed a significant increase in C3aR signal on microglia in the hippocampal CA1 region in comparison to the CA3 or DG region at 7 DPO (Fig. 4B and S5A-5C; t = 2.9, P = 0.0441; t = 1.730, P = 0.1587; t = 1.715, P = 0.1615; respectively). Neurons showed almost no co-localization with C3aR (Figure S5D). Next, we examined the modulatory effect of the C3-C3aR axis on microglial phagocytic function. Application of recombinant C3 to cultured primary microglial cells increased the number of fluorescent latex beads engulfed by microglia, while pretreatment with a C3aR antagonist partially blocked this effect (Fig. 4C and E; F (3, 30) = 19.33, P < 0.0001; F (3, 429) = 24.9, P < 0.0001; respectively). Moreover, blockade of C3aR inhibited both microglial activation and excitatory synapse loss after surgery (Fig. 4F and H; F (3, 8) = 31.53, P < 0.0001; F (3, 8) = 16.88, P = 0.0008; F (3, 20) = 9.992, P = 0.0003; respectively). Furthermore, we investigated the impact of C3aR inhibition on excitatory synaptic transmission utilizing whole-cell patch-clamp recordings. A remarkable reduction in the spontaneous excitatory synaptic transmission (sEPSC) frequency was observed in the mouse model. However, intraperitoneal injection of C3aR antagonists largely prevented the reduction in the sEPSC frequency in hippocampal pyramidal neurons of the mouse model (Fig. 4I, J and F (3, 64) = 4.073, P = 0.0104; F (3, 64) = 1.322, P = 0.2749; respectively). Together, these results strongly suggest that the microglial C3-C3aR axis participates in excitatory synaptic pruning in the mouse model.

Fig. 4

C3aR antagonist attenuates microglial activation and excitatory synapse loss in the mouse model of POCD. (A) Heatmap showing the upregulation of complement receptor genes in mouse hippocampus after peripheral surgery (n = 4–5 mice per group, each column represents an individual mouse). (B) Immunofluorescence detection of IBA1 (green) & C3aR (red) colocalized puncta in the hippocampus of the control and surgery group (n = 3 mice per group). Scale bar: 40 μm. (C-E) The effect of C3 and C3aR antagonist on bead phagocytosis activity of primary microglia cells. (C) Diagram showing the timeline of C3/C3aRA treatment and beads phagocytosis assay in primary microglial cells. (D) IBA1 & fluorescent beads double-labeling in primary microglial cells. Scale bar: 40 μm and 9 μm. (E) Quantification of the fluorescent beads taken up by primary microglial cells from (D) (left: 8–9 fields from 3 batches per group; right: 89–123 cells from 3 batches per group). (F-H) The effect of C3aR antagonist on microglial phagocytosis of excitatory synapses in vivo. (F) IBA1 (green) & CD68 (red) double-staining of hippocampal slices from control or surgery mouse treated with C3aRA or vehicle control. Scale bar: 100 μm and 20 μm. (G) Quantification of IBA1 or CD68 immunofluorescence intensity from (F) (n = 3 mice per group). (H) Immunoblot analysis of VGluT1 expression in control or surgery mouse treated with C3aRA or vehicle control (n = 6 mice per group). (I-J) The effect of C3aR antagonist on excitatory synaptic transmission in vivo. (I) Representative whole-cell patch-clamp recordings showing the spontaneous excitatory synaptic transmission (sEPSC) of hippocampal pyramidal neurons from control or surgery mice after vehicle or C3aRA treatment. (J) Quantification of sEPSC frequency and amplitude from (I) (n = 15–19 cells from 3 to 4 mice per group). Data in bar graph are presented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by two-tailed t-test or one-way ANOVA followed by Tukey’s post hoc test

C3aR antagonist rescues dendritic spine loss and LTP induction, further ameliorating memory dysfunction in the mouse model

To ascertained the impact of C3aR repression on LTP induction in the mouse model. We first measured the hippocampal CA1 area in the context of the dendritic spine density after surgery. Golgi staining exhibited that the total dendritic length and intersection number were not significantly altered in THE model mice compared to control mice (Fig. 5A, B and F (3, 34) = 0.6744, P = 0.5737; F (3, 314) = 1.194, P = 0.312; respectively). However, dendritic spines, in terms of their number, were remarkably decreased in CA1 neurons after surgery, and this effect was largely rescued by the administration of a C3aR antagonist (C3aRA) (Fig. 5C, D and F (3, 80) = 15.02, P < 0.0001). We next examined the impact of C3aR suppression on neuronal synaptic plasticity in the mouse model (Fig. 5E and F). LTP induction was drastically decreased in mice that underwent surgery compared with control mice, while the administration of C3aRA significantly prevented this loss (Fig. 5G, H and F (3, 40) = 7.412, P = 0.0005). Next, we investigated the function of C3aR in memory regulation in the model mice using the Y-maze test and LMT. As expected, mice that underwent surgery exhibited obvious memory deficits, as indicated by lower alternation percentage and poorer spatial recognition. Intriguingly, the administration of C3aRA almost completely reversed memory dysfunction in the model mice (Fig. 5I, J and F (3, 38) = 3.914, P = 0.0157; F (1, 76) = 18.26, P < 0.0001; respectively). These results infer that blockade of C3aR exerts a protective effect on dendritic spine density, synaptic plasticity, and memory after surgery.

Fig. 5

C3aR inhibition rescues the loss of dendritic spines and the impairment of LTP induction, and improves memory dysfunction in the mouse model. (A-D) Effect of C3aR antagonist on the loss of dendritic spines in POCD model. (A) Representative Golgi staining images of hippocampal pyramidal neurons from control or surgery mice after vehicle or C3aRA treatment. Scale bar: 50 μm. (B) Quantification of total dendritic length and intersection numbers of pyramidal neurons in (A) (n = 9–11 neurons from 3 to 4 mice per group). (C and D) Representative images (C) and quantification (D) of the dendritic spine density of hippocampal pyramidal neurons from control or surgery mice after vehicle or C3aRA treatment (n = 21 neurons from 3 to 4 mice per group). (E-H) Effect of C3aR antagonist on LTP induction in the model. (E) Representative image showing the location of stimulating (left) and recording (left) electrode in mouse hippocampal slice. (F) Example showing the field excitatory postsynaptic potential (fEPSP) trace. (G and H) Images (G) and quantification (H) of fEPSP in response to theta burst stimulation (delivering to the Schaffer collateral pathway) of hippocampal slices from control or surgery mice after vehicle or C3aRA treatment (n = 10–11 slices from 3 to 4 mice per group). (I-J) Effect of C3aR antagonist on memory deficits in the model. (I) Y-maze test showing the working memory performance in control or surgery after vehicle or C3aRA treatment (n = 9–12 mice per group). (J) LMT test showing the location memory performance in control or surgery mice after vehicle or C3aRA treatment (n = 9–12 mice per group). Data in bar graph are presented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by two-tailed t-test, one-way or two-way ANOVA followed by Tukey’s post hoc test

Genetic knockout of C3aR prevents excitatory synapse loss and memory deficits in mice after surgery

To confirm the essential function of C3aR in microglial-mediated excitatory synapse loss in the POCD model, we first downregulated the expression level of C3aR in the BV2 microglia cell line using lentivirus-delivered shC3ar1 (LV-shC3ar1) and evaluated the effect on the phagocytic ability of BV2 cells in vitro (Fig. 6A and B). The knockdown efficiency was validated using qPCR and western blot analysis (Fig. 6C, D and t = 6.448, P = 0.003; t = 2.455, P = 0.0494; respectively). Interestingly, the knockdown of C3aR partially abolished the C3-elicited phagocytosis activity (Fig. 6E, F and F (3, 32) = 37.86, P < 0.0001). Next, to ascertain the involvement of C3aR in excitatory synapse dynamics in vivo, we established a POCD animal model using C3ar1^−/− mice and compared the degree of excitatory synapse loss with that of wild-type mice. As expected, C3ar1 knockout markedly inhibited microglia phagocytosis after surgery, as indicated by the less enlarged volume of microglia and lysosomes compared with wild-type mice. Moreover, the volume of VGluT1-positive puncta within microglial lysosomes was much smaller in C3ar1^−/− mice than in control mice (Fig. 6G, H and t = 5.294, P < 0.0001; t = 4.588, P < 0.0001; t = 2.761, P = 0.009; respectively). In addition, VGluT1 expression levels were significantly elevated in the hippocampi of C3ar1^−/− mice postoperation (Fig. 6I, J and t = 3.538, P = 0.0241). The above findings verified that the C3-C3aR axis functions in excitatory synapse elimination in the mouse model. Furthermore, C3ar1^−/− mice that underwent surgery exhibited better memory performance than wild-type mice, as indicated by a higher alternation percentage in the Y-maze test; however, the total distance covered or traveled was not substantially different between the groups (Fig. 6K, L and t = 0.5446, P = 0.596; t = 4.644, P = 0.0006; respectively). Together, these results suggest that C3aR deficiency can inhibit microglial phagocytosis, prevent excitatory synapse loss, and alleviate memory deficits in the POCD mouse model.

Fig. 6

Genetic knockout of C3ar1 inhibits microglia-mediated excitatory synapse loss and memory deficits in the mouse model. (A) Schematic showing the C3ar1-shRNA lentivirus construct. (B) Representative image showing the virus transfection efficiency in the BV2 cell line. Scale bar: 100 μm. (C) Quantification of C3ar1 RNA level in BV2 cells infected with shCon or shC3ar1 using qPCR (n = 3 batches per group). (D) Immunoblot analysis of C3aR expression level in BV2 cells infected with shCon or shC3ar1 (n = 4 batches per group). (E-F) Effect of C3ar1 knockdown on the phagocytic activity of BV2 cells. (E) Diagram showing the timepoints of C3ar1 knockdown, C3 treatment and fluorescent bead phagocytosis assay in the BV2 cell line. (F) Representative images (left) and quantification results (right) showing the number of beads (red) engulfed by BV2 cells (n = 9 fields from 3 batches per group). Scale bar: 100 μm and 20 μm. (G-J) Effect of C3ar1 knockout on the phagocytic activity of microglia cells in POCD model mice. (G) 3D reconstruction of IBA1 (green), CD68 (blue), and VGluT1 (red) in the hippocampus of WT or C3ar1^−/− mice after surgery. Scale bar: 10 μm and 5 μm. (H) Quantification of microglial volume, CD68 volume, and VGluT1 positive puncta volume within microglial lysosome in the hippocampus of WT or C3ar1^−/− mice after surgery (n = 19 cells from 3 to 4 mice per group). (I) Immunoblot analysis of VGluT1 expression level in the hippocampus of WT or C3ar1^−/− mice after surgery. (J) Quantification of VGluT1 protein level in (I) (n = 3 mice per group). (K-L) Effect of C3ar1 knockout on the memory function of POCD model mice. (K) Open-field test showing the traveled distance in the hippocampus of WT or C3ar1^−/− mice after surgery (n = 7 mice per group). (L) Y-maze test showing the working memory performance in the hippocampus of WT or C3ar1^−/− mice after surgery (n = 7 mice per group). Data in bar graphs are presented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by two-tailed t-test or one-way ANOVA followed by Tukey’s post hoc test

Inhibition of p-STAT3 activation rescues excitatory synapse loss and memory deficits in mice after surgery

To explore the potential downstream pathways of C3aR, we found a direct link between C3ar1 and stat3 through transcriptomic analysis (Fig. 7A). Interestingly, the level of p-STAT3 expression was significantly increased in the mouse hippocampus after surgery and partially downregulated by C3aRA (Fig. 7B, C and F (3, 8) = 6.904, P = 0.0131; F (3, 8) = 7.429, P = 0.0106; respectively). Similarly, in primary cultured microglia cells, the application of C3 increased STAT3 phosphorylation, whereas pretreatment with C3aRA blocked this effect (Fig. 7D, E and F (3, 33) = 12.61, P < 0.0001). These results suggest that p-STAT3 was a downstream target of C3/C3aR signaling and was significantly upregulated after surgery. We therefore reasoned that inhibiting p-STAT3 signaling may have an effect similar to that of blocking C3aR. To prove this, we used stattic (a STAT3 phosphorylation inhibitor) in primary microglia culture and found that it could inhibit the elevation of p-STAT3 expression induced by C3 treatment (Fig. 7F and G; F (3, 32) = 12.94, P < 0.0001) and reduce the phagocytic ability of primary microglia (Fig. 7H and I; F (3, 32) = 9.758, P = 0.0001; F (3, 726) = 13.02, P < 0.0001; respectively). In vivo, stattic also effectively inhibited microglial activation (Fig. 7J, K and F (3, 12) = 44.27, P < 0.0001; F (3, 12) = 18.91, P < 0.0001; respectively) and prevented the loss of VGluT1-positive puncta at 7 DPO (Fig. 7L and M; F (3, 12) = 7.462, P = 0.0044). Finally, we evaluated the effect of p-STAT3 signaling inhibition on memory deficits in the mouse model. Notably, administration of stattic significantly increased the successful alternation percentage of THE model mice in the Y-maze test without affecting the distance traveled (Fig. 7N, O and F (3, 27) = 20.83, P < 0.0001; F (3, 27) = 8.802, P = 0.0003; respectively). Taken together, these results suggest that blockade of p-STAT3 signaling inhibits microglial phagocytosis and alleviates memory deficits in the mouse model.

Fig. 7

p-STAT3 inhibition attenuates microglia activation and improves memory impairment in the mouse model. (A) Gene network analysis showing the upregulated genes associated with the hub genes C3ar1 and Stat3. (B and C) Effect of C3aR antagonist on STAT3 and p-STAT3 expression in the model. (B) Representative immunoblots of p-STAT3 and STAT3 in the hippocampus of control or surgery mice after vehicle or C3aRA treatment (n = 3 mice per group). (C) Quantification of immunoblots in (B). (D and E) Immunofluorescence detection of p-STAT3 (green) in primary microglial cells after treatment with C3, C3aRA, or both (n = 9–10 fields from 3 batches per group). Scale bar: 20 μm. (F-G) Immunofluorescence detection of p-STAT3 (red) in primary microglial cells after treatment with C3, stattic (p-STAT3 inhibitor), or both (n = 9 fields from 3 batches per group). Scale bar: 100 μm and 2 μm. (H and I) Effect of p-STAT3 inhibition on the phagocytic activity of primary microglia cells. (H) Immunofluorescence detection of the number of beads (green) phagocytized by IBA1+ (red) primary microglial cells after C3 or stattic treatment. Scale bar: 100 μm and 10 μm. (I) Quantification of (H) (top: 9 fields from 3 batches per group; bottom: 130–286 cells from 3 batches per group). (J and K) Effect of p-STAT3 inhibition on microglia activation status in the model. (J) IBA1(green) & CD68 (red) double-labeling of hippocampal slices from control or surgery mice after vehicle or stattic treatment. Scale bar: 100 μm and 20 μm. (K) Quantification of (J) (n = 4 mice per group). (L and M) Effect of p-STAT3 inhibition on excitatory synapses in the model. (L) Immunoblott analysis of VGluT1 expression level in the hippocampus of control or surgery mice after vehicle or stattic treatment. (M) Quantification of (L) (n = 4 mice per group). (N-O) Effect of p-STAT3 inhibition on memory function in the model. (N) Open-field test showing the traveled distance in control or surgery mice after vehicle or stattic treatment (n = 7–8 mice per group). (O) Y-maze test showing the working memory performance in control or surgery mice after vehicle or stattic treatment (n = 7–8 mice per group). Data in bar graph are presented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns, not significant; by one-way ANOVA followed by Tukey’s post hoc test

Discussion

Excitatory synapse loss associated with memory dysfunction has been reported in multiple neurodegenerative disorders. In this study, we uncovered a driving role for microglia in excitatory synapse elimination and aggravation of memory deficits in the mouse model. Activation of the C3-C3aR axis was affirmed to exhibit an essential function in microglia-mediated phagocytosis of excitatory synapses and memory dysfunction. In contrast, knockout of C3aR or pharmacological inhibition of C3aR or the downstream STAT3 prevented excitatory synapse loss and memory dysfunction, indicating that complement signaling could be a potential therapeutic target for alleviating memory deficits in POCD.

Anesthesia and surgery are well-documented forms of trauma that can induce neuroinflammation in the brain [6, 33, 34]. As a common pathological feature associated with neuroinflammation, gliosis in the hippocampus has been documented to be the cause of cognitive impairment in the model mice [10, 11, 35, 36]. Microglia functions in the regulation of CNS homeostasis as well as neuroinflammation. We observed that surgery directly induced activated response microglia (ARM) in the model mice [37], which is responsible for the loss of excitatory synapses, abnormal synaptic transmission, and learning and memory deficits. Although many pieces of evidence have demonstrated that the complement system functions in synapse elimination and memory deficits in the developing brain and several neurodegenerative disorders such as AD [38, 39], the role of the complement cascade in mediating synapse elimination in the model mice remain controversial. A recent study reported that complement C1q but not C3 promotes postsynaptic element engulfment by microglia after extended anesthesia induction [40]. In contrast, our results suggested that the expression level of complement C3 but not C1q was elevated after brief anesthesia exposure in mice that underwent surgery. Furthermore, a large amount of complement C3 but not C1q is colocalized with excitatory presynaptic structures, indicating that more excitatory synapses could be engulfed by microglia in our mouse model. We infer that the differences between our studies and other works are largely attributed to the variation in the mouse model used. Anesthetics, which are neurotoxic drugs, may directly induce glial cell activation and synapse loss after long-term exposure [40, 41]. In our mouse model, we reasoned that surgery was the most important trigger of microglial activation and excitatory synapse loss. Similar to our findings, another study also revealed abundant glial cell proliferation associated with complement activation in the mouse hippocampus after peripheral surgery [19].

Elevated levels of complement proteins and the activation of their signaling pathways have been reported in many diseases. However, there is still controversy regarding the specific cell types that express the C3/C3aR pathway proteins. Our immunofluorescence experiments revealed that C3 is primarily derived from astrocytes and is minimally expressed in microglia and neurons, while C3aR is mainly expressed in microglia and rarely in neurons (Figure S5D). Both of these proteins showed a significant increase in expression in the hippocampus of the model mice subjected to surgical procedures. These results are consistent with recent studies [22, 26]. Interestingly, in our model, although there was a trend of increased expression of C3 and C3aR in the CA3 and dentate gyrus regions of the hippocampus, the increase in the CA1 region was statistically significant. The hippocampus, being a brain region closely associated with cognition, may have different subregions that play distinct roles in certain functions. The CA1 region is a crucial area for communication between the hippocampus and the cortex. The latest study [42] have shown that, compared to wild-type mice, the complex neuronal interactions in the dorsal CA1 hippocampal neurons of amnestic mice are disrupted. This suggest that the CA1 region of the hippocampus may be a susceptible brain area for cognitive-related impairments. In our surgery-induced cognitive impairment model, the changes in C3/C3aR and neuronal functional damage in the CA1 region of the hippocampus are more pronounced, which seems to validate this point.

Pharmacological inhibition or genetic knockout of C3aR can effectively alleviate neuroinflammation and ameliorate memory impairment in AD mice or virus infection mouse models [26, 43]. Similarly, C3aR inhibition or genetic knockout prevented the loss of excitatory synapses and the reduction in excitatory synaptic transmission and rescued memory impairment in the mouse model. The STAT3/p-STAT3 pathway has been reported to be an important downstream signaling pathway underlying C3aR activation. The activation of C3aR leads to the phosphorylation and activation of STAT3. Activated STAT3 translocate to the nucleus, where it can induce the expression of genes involved in immune responses and cellular functions. Activation of STAT3 in microglia promotes their transformation into ARM, which perform various functions including synaptic pruning. VGluT1 (vesicular glutamate transporter 1) is a marker for excitatory synaptic vesicles. The activated microglia recognize and engulf VGluT1-positive synapses, effectively pruning these excitatory synapses. Although we excluded the possibility of extensive cell death through TUNEL staining, we still cannot completely rule out the possibility that the engulfed VGluT1 comes from naturally apoptotic cell debris. Ideally, it would be more effective to use two-photon microscopy to observe the real-time dynamic interactions between microglia and neurons. This would help clearly distinguish whether microglia are engulfing intact excitatory synapses or some components released into the extracellular matrix. Nevertheless, our experimental observations, combined with omics data analysis, can provide some insight. They suggest that activated microglia may be engulfing more synaptic components, which in turn affects the normal function of synapses, resulting in cognitive impairment. Together, pharmacological and genetic evidence demonstrates the central role of the C3aR/STAT3 signaling axis in microglia activation, excitatory synapse loss, and memory deficits in the model mice.

There are still some questions remaining to be answered in future work. First, although C3aR has been confirmed to be primarily expressed in microglia in the central nervous system, C3aR knockout mice do not represent a conditional knockout in microglia and cannot completely rule out the effects of C3aR knockout on other neural cells. Ideally, using microglia-conditional C3aR knockout mice would be more persuasive. Therefore, in future studies on microglia, we will incorporate Cx3cr1^cre mice. Second, the present study revealed that microglia participate in excitatory synapse elimination, whether microglia are also involved in inhibitory synaptic pruning and their influence on memory deficits in POCD model mice remain to be addressed. Third, as another important glial cell type, astrocytes are also involved in synapse development and pruning. Therefore, the function of astrocytes in excitatory synapse elimination in POCD model mice is worthy of investigation.

In summary, using a POCD mouse model, we discovered that the C3-C3aR/STAT3 axis was significantly activated and mediated microglial-neuronal crosstalk in the mouse hippocampus after surgery. Blocking the C3-C3aR axis and downstream STAT3 signaling inhibited surgery-induced excitatory synapse loss in the hippocampus and relieved memory deficits in POCD model mice. Therefore, targeting the complement system may be a potential therapeutic strategy for alleviating memory deficits in elderly patients after surgery.